mediaeval thought-experiments: the metamethodology...

c© Peter King, in Thought-Experiments (Rowman & Littlefield 1991), 43–64.

MEDIAEVAL THOUGHT-EXPERIMENTS:

The Metamethodology of Mediæval Science

1. Introduction

The modern understanding of mediæval science has been shaped by thepioneering scholarship of Pierre Duhem, who put forward as a central claimthat the components of “modern” science—most notably the kinematic anddynamic insights of Galileo—are found, or at least are prefigured, in earliermediæval scientific writings, out of which modern science evolved.1 Scholarswho have followed in Duhem’s footsteps, such as Charles H. Haskins, An-naliese Maier, Edward Grant, A. C. Crombie, William A. Wallace, MarshallClagett, James A. Weisheipl, Lynn Thorndike, and John Murdoch, havecarried out their research into mediæval science in the shadow of Duhem’scentral claim. Dissatisfaction is occasionally expressed, but, in the absenceof a developed alternative to Duhemian orthodoxy, such complaints echo ina void, without any theoretical support.2

1 See Pierre Duhem, Etudes sur Leonard de Vinci: ceux qu’il a lu, ceux qui l’ont lu, 3

vols., Paris 1906–1913, and Le systeme du monde: histoire des doctrines cosmologiques

de Platon a Copernic, 10 vols., Paris 1913–1959. He concisely summarized his viewsin the Rendiconti della Reale Accademica dei Lincei (Classe scientia, fisica, matemat-

ica) 22 (1913), 429: “In the fourteenth century, the Masters of Paris, having rebelled

against Aristotle’s authority, constructed a dynamics entirely different from that ofthe Stagirite; the essential elements of the principles thought to have received math-

ematical expression and experimental confirmation from Galileo and Descartes were

already contained in this dynamics. . . Galileo, in his youth, read several of the trea-tises where these theories were presented.” Marshall Clagett, echoing these sentiments,

has written that “Galileo was clearly the heir to medieval kinematics” (The Science

of Mechanics in the Middle Ages, Wisconsin 1959, 666, hereafter abbreviated as SM).Research into Galileo’s knowledge of mediæval science continues; see A. C. Crombie,

“Sources of Galileo’s Early Natural Philosophy” in Reason, Experiment, and Mysti-cism in the Scientific Revolution, edited by M. L. R. Bonelli and W. R. Shea, New

York 1975, and William A. Wallace, “Galileo Galilei and the Doctores Parisienses”

in New Perspectives on Galileo, edited by R. E. Butts and J. C. Pitt, New York1983. Duhem’s portrait of Galileo as a pure experimental physicist is undoubtedly

idealized—some would say mythologized—but it is the contrast between the method

of mediæval physics and modern experimental method which interests me, and ac-cordingly I will ignore the dubious historical accuracy of Duhem’s claims on behalf

of Galileo. When I speak of the “Duhemian tradition” I have in mind the works of

Duhem, Maier, Crombie, and Clagett, again prescinding from individual differences.2 John Murdoch, no believer in the Duhemian orthodoxy, has written: ”It would be

an error to regard these new and distinctive fourteenth-century efforts as moving

– 1 –

2 MEDIAEVAL THOUGHT-EXPERIMENTS

A reassessment of Duhem’s central claim presupposes an understandingof the mediæval paradigm of science formed independently of its possible in-tellectual progeny; once such an understanding has been attained, questionsof continuity and influence can be sensibly raised, but not before. In this ar-ticle, I will try to construct a picture of the scientific paradigm current in themid-fourteenth century, concentrating on physics for two reasons: Duhem’scase is strongest with regard to physics, and physics, unlike, say, alchemyor astrology, was recognized as a science (as scientia) at the time.3 Hence-forth when I speak of “mediæval science” my meaning is properly restrictedto fourteenth-century physics, unless noted otherwise. In particular, I willargue for five theses: (i) the achievements of mediæval science, even thethree achievements traditionally singled out as anticipating modern science,were part of a completely different scientific paradigm; (ii) this paradigmtook as the measure of success of its theories and hypotheses not experi-mental confirmation, empirical justification, or saving the appearances, butrather the ability to deal with examples and purported counterexamples;(iii) the method of mediæval science was thought-experiment rather than

very directly toward early modern science. . . the whole enterprise of many a medievalscholar who treated motions was worlds away from that of Galileo and his confreres. . .medieval discussions of motion should not be viewed solely as providing some kind ofbackground from, or against which, early modern thinking about motion developed”

(John Murdoch and Edith Sylla, “The Science of Motion,” in Science in the MiddleAges, edited by David Lindberg, Chicago 1978). This sentiment seems to me to be

exactly correct, lacking only a positive alternative conception of mediæval physics.3 The restriction is non-trivial. There was a great deal of practical knowledge, and

even theoretical knowledge, about scientific matters during the Middle Ages whichwas not recognized as scientia and hence not subject to pressures that it conform

to a paradigm, for example alchemy and pharmacology. My argument is directed atthe academic conception of scientia current in the fourteenth century, discussed in §3below. Astronomy provides a good example of the kind of split I have in mind: as

an academic discipline, astronomy was classified as a branch of mathematics, primar-ily concerned with what we should call celestial mechanics; its debates were largely

geometrical, working out the ramifications of the theoretical system of epicycles or

concentric spheres. Such theories only took into account very general phenomena asconstraints and were rarely tested against experience. However, astronomical mat-

ters were connected to practice in three ways: for purposes of navigation, whether

exploratory, commercial, or military; for calculating the date of religious feasts; forthe popular though officially disapproved uses of astrology. Each of these practical

needs spurred the collection of observations, keeping of records, adoption and im-

provement of technology (such as the astrolabe and the sextant), and collectivelyaccounted for whatever scientific progress there was in mediæval astronomy. To some

extent, the mediæval science of optics (scientia perspectiva) is similar; it too was clas-sified as a branch of mechanics, largely concerned with trigonometry, but had practical

connections.


2. THREE ACHIEVEMENTS OF MEDIAEVAL SCIENCE 3

actual experiment or testing; (iv) there was a developed body of reflectionon the method of thought-experiment, found in treatises on obligationes,which constituted the meta-methodology or philosophy of scientific methodin support of mediæval scientific practice; and (v) this method has its ownvirtues and vices quite distinct from those of modern scientific method.

2. Three Achievements of Mediaeval Science

The Duhemian tradition identifies three achievements of mediæval phy-sics, all put forth in the second quarter of the fourteenth century, as sup-porting its central claim: Heytesbury’s Mean-Speed Theorem,4 Bradwar-dine’s Function,5 and Buridan’s theory of impetus.6 Each had an imme-

4 William Heytesbury (1305?-1372/3) states the Mean-Speed Theorem in his 1335 Reg-ulae solvendi sophismata (on fol. 40v of the 1494 Venice incunabulum of Heytesbury’s

works); a better version of the text in which the theorem is stated is given in Clagett,SM 277–283. No proof is offered in the Regulae, but a proof is offered in the Pro-bationes conclusionum, a separate treatise also attributed to Heytesbury appearing

in the 1494 Venice incunabulum. (The attribution to Heytesbury is not secure: seeCurtis Wilson, William Heytesbury: Mediæval Logic and the Rise of Mathematical

Physics, Wisconsin 1956, p.210, for a statement of the doubts.) The Mean-Speed

Theorem is also stated by two contemporaries of Heytesbury: Richard Swineshead(fl. 1340–1355), in the fourteenth chapter, entitled Regulae de motu locali, of his 1350

Liber calculationum (printed in 1477 Padua and 1520 Venice incunabula; Clagett

gives a better text, SM 298–304), states the theorem and offers four proofs; JohnDumbleton (fl. 1340–1350), in his Summa logicae et philosophiae naturalis (extant

only in manuscript, but text given in Clagett, SM 317–325), states the theorem and

offers a distinct indirect proof—one quoted verbatim by Giovanni Marliani in the latefifteenth century (Marshall Clagett, Giovanni Marliani and Late Medieval Physics,

New York 1941). Swineshead and Dumbleton were, like Heytesbury, members of Mer-

ton College at Oxford; in consequence, the theorem is sometimes called the ”MertonMean-Speed Theorem.” It may, however, be an exaggeration to think there was a

distinct school of thought, the ‘Mertonian School,’ to which these thinkers, along withRichard Kilvington and Thomas Bradwardine (and Walter Burleigh as an honorary

member), belonged; see Edith Sylla, “The Oxford Calculators” in The Cambridge His-

tory of Later Mediæval Philosophy, eds. N. Kretzmann, A. Kenny, and J. Pinborg,Cambridge 1982, 540–563.

5 Bradwardine (1295?–1349) states his general rule governing the ratio of proportions,

which has come to be known as “Bradwardine’s Function,” in his 1328 Tractatus de

proportionibus velocitatum in motibus, edited and translated by H. Lamar Crosby,Thomas Bradwardine, His Tractatus de proportionibus: Its Significance for the De-

velopment of Medieval Physics, Wisconsin 1951, 112.6 Jean Buridan (fl. 1330–1360) argues for the theory of impetus in his Quaestiones sub-

tillissimae super octo Physicorum libros Aristotelis VIII q. 12, printed in a 1509 Parisincunabulum and reprinted as Johannes Buridanus: Kommentar zur Aristotelischen

Physik, Minerva G.M.B.H. Frankfurt-am-Main 1964; other discussions by Buridan



diate influence on contemporaries; each, it is claimed, prefigures Galileo’sscientific investigations; each is said to be directly influential in the historyof physics—to such an extent that Galileo is accused of pirating mediævalresults.• The Mean-Speed Theorem. In cases of bodies moving with constant ac-

celeration (positive or negative), William Heytesbury argued that thedistance traversed would be the same as if the body were moving for thesame length of time and no acceleration with a uniform speed7 equal tothe instantaneous speed at the middle instant of time of the interval ofuniform acceleration.

• Bradwardine’s Function. To determine the speed at which a body with agiven force will move through a medium with a given resistance, ThomasBradwardine argued for the general claim that the speed will vary arith-metically when the proportions of force to resistance are varied geomet-rically.

• The Theory of Impetus. To explain the motion of bodies not in obviouscontact with any mover, as in the case of projectile motion, Jean Buridanargued for the existence of an enduring quality, called impetus, impartedto the moving body by the mover, which varies in direct proportion tothe mass (the quantity of prime matter) and the speed of the movingbody.

The claims made on behalf of these three achievements of mediæval physicsin the Duhemian tradition are forceful and striking. Heytesbury’s Mean-Speed Theorem has been cited as an anticipation of Galileo’s analysis ofacceleration of bodies in free fall, Duhem suggesting that Galileo was ac-quainted with it.8 Bradwardine’s Function has been considered a mathe-

are thoroughly explored in Annaliese Maier, Zwei Grundprobleme der scholastischenNaturphilosophie, 3rd edition, Rome 1969, 101–128.

7 I regularly translate velocitas by ‘speed’ rather than ‘velocity’, since modern physicsconstrues velocity as a vector, and speed as a scalar, quantity. The claim of Marshall

Clagett, “Richard Swineshead and Late Medieval Physics,” Osiris 9 (1950), 131–

161, that Swineshead in fact distinguishes between the quantity of motion and thedirection of motion, seem to me a strained reading of the text, and in any event is not

sufficient to show that one and the same thing, velocitas, possessed both magnitude

and direction (i. e. is a vector).

8 The mathematical formalization of the Mean-Speed Theorem, taking d as distance

over the interval [a, b], S as speed (evaluated at the extrema of the interval), and t astime, is: d ∝ (Sa + (Sb − Sa)/2)t. Since the difference Sb − Sa = at (where a is the

acceleration), this reduces to the familiar kinematic equation d ∝ Sa + at2/2. Wherethe initial speed is zero, as in the case of dropping a body, the result is d ∝ at2/2,

that is, the distance traversed is proportional to the square of the time elapsed, which



matical breakthrough to the theory of exponential variation, and in its anal-ysis of force, resistance, and speed likened to Galileo’s analysis of these.9

With regard to Buridan, Clagett writes: “one cannot help but compareBuridan’s impetus with Galileo’s impeto, and Newton’s quantity of motion(momentum).”10 As well as forceful and striking, these claims may seemextravagant. And so they are.

The most obvious grounds for skepticism about such claims come fromthe material basis of mediæval physics: rather than the result of directobservation and experimentation, the enterprise of mediæval physics is tex-tual, either commentary or questions on Aristotle, or independent treatiseson particular questions derived from Aristotle. Heytesbury states the Mean-Speed Theorem in his Regulae solvendi sophismata: a book for intermedi-ate logic students, giving rules for resolving logical puzzles (sophisms).11 Asurvey of its six chapters should point up the skeptical doubts. The first,“Insolubles,” is concerned with the semantic Paradox of the Liar and its vari-

may be applied directly to the case of free fall. However, the first reference to the free-fall application of the Mean-Speed Theorem appears only in Domingo de Soto’s 1555

Quaestiones super octo libros Physicorum Aristotelis (Salamanca 1555 incunabulum,

fol. 92v), not in the High Middle Ages. See Pierre Duhem, Le systeme du monde,Vol. 3 128ff.

9 Mediæval mathematics was not well-equipped to deal with exponential variation. Mar-

shall Clagett, in SM 675, says of Bradwardine’s Function that “it foreshadowed the

differential equation used so universally in modern mechanics.” Annaliese Maier, inAusgehendes Mittelalter Vol. I, Rome 1964, 425–457, is more emphatic: “Without

a doubt it was Thomas Bradwardine who. . . anticipated to the greatest extent themethodology of the natural science of the future.”

10 Marshall Clagett, SM 523; on 525 he says “it can scarcely be doubted that impetusis analogous to [Newton’s] inertia.” This sentiment is representative. See also, for

example, Pierre Duhem, Le systeme du monde Vol. 2 154; Alexandre Koyre, Etudes

Galileennes, 3 vols. (Actualites Scientifiques et Industrielles, Nos. 852–54), Paris 1939,Vol. 3 113–142; Annaliese Maier, Die Vorlaufer Galileis im 14. Jahrhundert, Rome

1949, 137ff. The comparison to Galileo is based on the content of Galileo’s claimsabout impeto, while the comparison to Newton is made plausible due to Buridan’s

loose formulation: if impetus varies in proportion to the quantity of matter and varies

in proportion to the speed, then it seems to follow that it varies in proportion to theproduct of the quantity of the matter and the speed—which is taken as the analogue

of the Newtonian analysis of momentum as the product of mass and velocity. (The

comparison with Newtonian inertia is also supported on the strength of Buridan’sclaim that it is an enduring quality.)

11 James Weisheipl, in “Ockham and Some Mertonians,” Mediæval Studies 30, 196–

197, has expressed doubts about Heytesbury’s veracity given the high level of logical

sophistication displayed in the Regulae, but Heytesbury is quite explicit on this point:vestrae sollicitudini iuvenes studio logicalium agentes annum primum prout facultatis

meae administraret sterilitas (on fol. 4va of the 1494 Venice incunabulum).



ations. The second, “Knowing and Doubting,” deals with what we shouldcall epistemic-doxastic logic, investigating the logical behavior of proposi-tions with such terms. The third, “Relative-Terms,” deals with questionsof what we should now call anaphoric reference. The fourth, “Beginningand Ceasing,” deals with the assignment of intervals and limit-points of in-tervals (e. g. “Socrates begins to run”). The fifth, “Maxima and Minima,”deals with the extrema of functional variation and the conditions underwhich such extrema exist. The sixth and last chapter deals with the threecategories in which Aristotle says that non-substantial change takes place:Quantity (growth or augmentation), Quality (alteration), and Place (loco-motion). The Mean-Speed Theorem is stated in this last section. The entirecontext of the Regulae is logical, or logico-mathematical; it is far removedfrom the experimental researches of a Galileo, or even of a contemporaryalchemist. Certainly, the literary form of a work need not reflect its content,but in this case I think it tells against the experimental nature of mediævalphysics: the conception of ‘physics’ under which the Mean-Speed Theoremis naturally classified together with anaphora and paradoxes of self-referenceis distinctly non-modern, having little to do with experimental confirmationof hypotheses.

A sense of the alternatives available at the time sharpens the doubt.Heytesbury’s slightly older contemporary, Richard Kilvington, addressesthe question of instantaneous speed in his 1325 Sophismata, rejecting theMean-Speed Theorem for two main reasons.12 First, because it begs thequestion; one must stipulate the instantaneous speed to determine the in-stantaneous speed. Second, because it is conceptually incoherent; ‘speed’is the rate of motion, but motion only takes place during an interval andnot at any instant. Instead, Kilvington maintains that there is no senseto “instantaneous speed” at all.13 How were such debates as these set-

12 An edition of Kilvington’s Sophismata is being prepared by Norman and BarbaraKretzmann; my information is derived from a typescript of their work, for which I am

grateful. Kilvington attacks the notion of instantaneous speed in Sophisms 33/34.13 A modern way to put his point would be to say that S = d/t, which is undefined

for t = 0, i. e. over no interval. This complaint is echoed by others. Marsilius of

Inghen, author of the Quaestiones in octo libros Physicorum Aristotelis previouslyattributed to Duns Scotus (and appearing in the Wadding-Vives edition of his works),

says in Book III q. 5 “instantaneous motion is neither fast (velox) nor slow [i. e. has

no ‘speed’ at all], since these are only defined with respect to [an interval of] time.”One of Kilvington’s arguments suggests taking an arbitrarily small interval, where the

point at which the speed is to be determined is intrinsic to the interval; no matterhow small the interval chosen, the absolute value of the difference of the functional

extrema is nonzero; and so, Kilvington concludes, the notion of ‘instantaneous speed’



tled? Whose position was better? And the answer is: by recourse to logicalconsiderations—example, counterexample, generalizing plausible rules cov-ering some cases to fit others. There is no hint of anything empirical in thisprocess.

Now it could be, and indeed has been, maintained that Heytesburyand Kilvington are working on highly abstract forms of physics, a kind ofmediæval “theoretical physics.” Support for this comes from the fact that itis not clear what would count as experimental or inductive support for theMean-Speed Theorem; it just isn’t the kind of thing which can be tested,there being no instruments capable of measuring speed at an instant. Itis, rather, a high-level conceptual innovation of mathematical physics. Fur-thermore, the logico-mathematical procedure is appropriate to determiningthe truth of the Mean-Speed Theorem, because it is peculiarly apt to re-veal conceptual weaknesses and expose hidden inconsistency or incoherence.Thus, it is said, mediæval physics is recognizably physical science, but at ahighly abstract level; not experimental, but theoretical, physics.

While there is some justice in this reply, I think it misses the point.Theoretical physics, no matter how theoretical, qualifies as physics for usin virtue of being hooked up to experimental testing in the long run, andthe Mean-Speed Theorem cannot be directly tied to experience. Is therean indirect connection? There is a testable consequence of the Mean-SpeedTheorem: the so-called “Distance Theorem.”14 Yet there is no indicationthat during the mediæval period the Distance Theorem was ever justified onany basis other than geometrical considerations. In fact, they could not haveperformed the relevant experiments: Galileo notes that he had to constructa more accurate clock in order to experimentally test it; the technologywould have been beyond mediæval resources. Just as it does not suffice for

is incoherent. If he had taken the difference in ratio to the length of the interval,he would be well on the way to the differential calculus, since mediæval mathematicspossessed the machinery of limit points and convergent sequences. Heytesbury does

not address the question.14 In the Probationes a special case of the Distance Theorem is proved on the basis of

the Mean-Speed Theorem, namely the claim that a body with constant acceleration

and beginning from rest will traverse three times as much distance in the second half

of the interval of time as in the first. The generalized form of the theorem, that inequal periods of time the body will traverse distances as proportionally related by

the series of odd numbers 1, 3, 5, 7, . . . was stated and proved by Nicole Oresme in his

Quaestiones in libro Euclidis Elementorum (MS Vaticana, Chis. F. IV.66 fol. 29rb, textgiven in Clagett, SM 344 n. 13). Galileo’s statement and discussion of the Distance

Theorem is given in his Discorsi e dimonstrazioni matematiche intorno a due nuovescienze (eds. A. Carugo and L. Geymonat, Turin 1958), at the beginning of the Third

Day (212 in the standard Edizione Nazionale system).



something to be called ‘physics’ that the content of its principles concernthe motion of bodies in space (for then astrology would qualify as physics),so too it does not suffice that the principles have testable consequences(astrology would again qualify).15 Rather, the justification of the physicalprinciple has to be tied to the actual testing, to actual experience. TheKilvington-Heytesbury debate, therefore, cannot be interpreted as a formof abstract, highly theoretical physics.

Bradwardine’s Function, although not universally accepted by those whowere studying natural philosophy, at least has the look of something whichwould be directly testable, unlike the Mean-Speed Theorem: all one need dois perform experiments on bodies moving with various forces through mediawith various resistances to see if the ratio of the speeds varies arithmeticallyas the ratio of force to resistance varies geometrically.† Yet Bradwardine didnot raise this question abstractly, to discover whether such relations obtain,nor for experimental reasons, to resolve a theoretical difficulty, but ratheras a textual problem: in Physics 7.5 249b27–250b7 Aristotle lists severalparticular examples of relations among speed, force, and resistance, which inthe Middle Ages were understood to be instances of an unstated general rule;the project was to find the rule. Many critics of Bradwardine’s Function—for example, Blaise of Parma, Giovanni Marliani, and Alessandro Achillini—rejected his appeal to a mathematical tradition that sharply distinguishedbetween fractions and proportions, and were unconcerned with its “fit” withexperience.

15 There is good science (analytic chemistry) and bad science (phlogiston theory); thereis also non-science (geometry) and pseudoscience (astrology). The distinction, no

matter how imprecise and difficult to draw, is nevertheless real. The Kilvington-

Heytesbury debate is non-science rather than bad science or pseudoscience; it is non-science because there are no links to experimental control.

† Annaliese Maier, in Metaphysische Hintergrunde der spatscholastischen Naturphiloso-phie, Rome 1955, 374–375, has mathematically formalized Bradwardine’s Function.

Her version it is non-arithmetic, intended to capture the notion of proportions varying

geometrically, but in fact is much stronger than Bradwardine’s actual claim. Geomet-ric proportions are squares, “doubling” the proportion, while Maier’s formalization

admits any nonzero real-valued exponent (any logarithmic base). Whereas modern

physics construes force as the product of mass and acceleration such that the forceexpended by a body moving through a medium is equal to that of the resistance of-

fered by the medium, the mediævals took “force” (vis or potentia) as a measure of thestrength of an active potency possessed by the moving body such that not all of the

potency need be actualized in moving through a medium. (Correlatively, “resistance”

was the measure of the strength of a passive potency possessed by the medium.) Hencethe mediæval notion of vis is closer to our notion of kinetic energy than our notion of

force.



Perhaps it is a good thing that Bradwardine was not concerned aboutexperimental confirmation of his function: the fact of the matter is thatan earlier theory, explicitly rejected by Bradwardine, seems to better fitexperience, namely the view that speed varies in proportion to the differencebetween force and resistance.16 The rejected theory is the natural one,given the mediæval understanding of force; Bradwardine’s rejection, like theproblem itself, is not based on experiment, but on textual considerations—he says that Aristotle (loc. cit.) takes force and resistance to be related by aproportion, not by their difference. And that, for Bradwardine, settles thematter.

The further development of Bradwardine’s Function, by philosopherssuch as Oresme and Swineshead, is mathematical rather than experimen-tal. The proportionality constants are never given; numerical values areassigned in examples for illustrative purposes, but not derived from testing.Heytesbury remarks that it would be more of a hindrance than a help to de-termine such values (Regulae, fol. 41rb), a sentiment echoed by Swineshead(Liber calculationum, fol. 52ra of the 1520 Venice incunabulum). Oresme’sextension of Bradwardine’s Function is used to criticize astrology, but thegrounds it proposes are solely mathematical, even though observational de-termination of the proportions would make the criticism decisive.17 Despiteits testable content, Bradwardine’s Function is no more a piece of modernphysics than the Mean-Speed Theorem.

16 Mathematically, the rejected theory holds that S ∝ F−R. Versions of this theory wereheld by earlier philosophers, most notably Philoponus, Avempace, Averroes, and later

by Galileo in his youth; see Ernest A. Moody, “Galileo and Avempace: The Dynamics

of the Leaning Tower Experiment,” Journal of the History of Ideas 12 (1951), 163–193and 375-422.

17 Nicole Oresme (1325?–1382), De proportionibus proportionum, edited by Edward

Grant, De proportionibus proportionum and Ad pauca respicientes, Wisconsin 1966,

384–387. Oresme develops the notion of ‘irrational’ proportions of proportions, i. e.equations in which (a/b)x = c/d only if x is irrational, arguing that it is mathemati-

cally probable that any two proportions chosen from a set of unknown proportions will

be irrational, the larger the set the greater the probability; hence the force-resistanceproportions governing the motion of celestial bodies are likely irrational, and so, by

Bradwardine’s function, their speeds are incommensurable, entailing that the celes-tial bodies never occupy the same position twice. But this last result contradicts the

basis of predictive astrology, which must therefore be unable to predict. In fairness

to Oresme, it should be noted that there was no way in fact to determine the propor-tionality constants required: there was no access to celestial phenomena, and there

were no instruments capable of such measurements.



3. Experience and Experiment

With Buridan the issues we have been discussing can be put in sharperfocus. When Buridan argues for the existence of impetus, he does seem toappeal to experience in a straightforward manner. Like Bradwardine, hebegins with a textual problem: Aristotle, in Physics 4.8 215a14–17, seemsto endorse a theory known as ‘antiperistasis’ to explain projectile motion,whereby the air surrounding the moving body rushes in from behind, inthe space vacated by the moving body, to impart a continuing force to thebody and keep it in motion. Buridan objects to this theory by citing three‘experiences,’ experientiae: (i) some things, such as a blacksmith’s wheel ora child’s top, spin without changing their place, and so it makes no senseto say that air rushes in from “behind” the moving body; (ii) reducing thesurface volume exposed to the air pushing from behind, as when the handleof a lance is sharpened to a point, does not reduce the projectile’s speed; (iii)if those pulling a boat upsteam suddenly stop, so that the boat continues todrift upstream through its momentum, a person on the boat does not feelwind from behind pushing him forward, and neither would straws at therear of the boat be bent over. On first glance, (i)–(iii) seem to be the sortof thing I was demanding: a recourse to experience to support or attacka theory. Indeed, that is how Buridan’s examples have been construed;Clagett praises their “strongly empirical and observational character.”18

But appearances can be deceptive and, in this case, are deceptive.With regard to the appeal to experience—to experientiae or to experi-

menta (Buridan uses both terms)—it is misleading to view such appealsas evidence for an empirical method. Often the sense of ‘experience’ is nostronger than the experience of the old farm hand, a simple way of stat-ing “what everybody knows.” Such appeals are not to be confused withmodern experiments or testing: no question is being put to Nature. Asmodern experiments, the standards of precision and care are laughable atbest. The recourse to commonly-held beliefs is not equivalent to observa-tional reports; common beliefs are theory-laden and held in the absenceof, and sometimes in the face of, evidence. What “everybody knows” is amiscellaneous mixture of fact, superstition, prejudice, report, rumor, andwishful thinking. Francis Bacon, well after the mediæval period, cited as acertain experimentum the ‘fact’ that a man whose feet are planted in thesoil at top of the Himalayas will live forever. Nor are there rules for whatcounts as relevant evidence; astrological predictions are “confirmed” time

18 Marshall Clagett, SM 563–564.


3. EXPERIENCE AND EXPERIMENT 11

and again by such unlikely events as a bird flying north on the day of a fatalaccident. The sense of “experience” Buridan appeals to supports astrologyas well as physics.

Even mundane examples of the sort cited by Buridan do not count asputting a question to Nature, since there is no record of anyone puttingsuch claims to the test. This is worth emphasizing: we have no record of amediæval physicist drawing a testable consequence from a theory and thenattempting to actually test it, or have it be tested. Buridan asserts thatsharpening the handle of a lance does not reduce its speed; why believehis claim? The answer that it need not be tested because it is “obvious”what would happen is misguided on two counts. First, the “obvious,” like“what everyone knows,” is treacherous. It was “obvious” that a large stonewill hit the ground before a small stone, if both are dropped simultaneouslyfrom the top of a tower. Second, and more important, this answer missesthe point, which is that Buridan’s appeal to experience is not an appeal toobservation, much less an appeal to observation in cases of intervention.19

Buridan does not seek to justify a physical theory, or to show that a physicaltheory is unjustified, on the basis of the evidence produced by controlledexperimentation; he appeals to common beliefs to illustrate his contentionthat a given theory is or is not justified. The experientiae or experimenta ofmediæval science are completely unlike the experiments of modern science.

This conclusion can be supported by considering the kinds of experien-tiae actually appealed to by mediæval physicists in addition to the casesmentioned above. Bradwardine rejects the theory that the speed of a bodyis proportional to the difference of the force and resistance, citing an exper-imentum in which a weight is attached to a fly. While discussing whethermotion is relative or absolute, Buridan posits that God could rotate theentire cosmos as a single solid body. To extend Bradwardine’s Function,Swineshead imagines a rod of uniform density approaching the center of theuniverse through a void. Such cases are introduced exactly on a par withthe mundane examples of sharpening lances—and this fact, if nothing else,should suggest that our understanding of the mediæval enterprise has to be

19 I use the term “intervention” as a catch-all to describe the characteristics of modern

scientific experimental method: the design of an experiment, including the separationand identification of the relevant variables; the use and development of technology to

bring about conditions for an experiment; the measure of reliability of the equipment;

the conditions for repeatability, including the control for generality in the sample, andactual repetition to the level required; the notion of statistical confirmation and statis-

tical projection of laws, entirely absent in the mediæval period; and other features—concerning which see the second part of Ian Hacking, Representing and Intervening:

Introductory Topics in the Philosophy of Natural Science, Cambridge 1983.



changed.If we focus on the wide variety of these cases for a moment, a general pic-

ture starts to emerge. The philosophers of the High Middle Ages, it seems,were generally indifferent to whether the examples they describe could beconcretely realized. It is not an essential feature of mediæval science that itproceed by way of realizable experientiae, much less by modern experiment.We could sharpen lances easily; we would have difficulty attaching a weightto a fly; we could not observe a rod approaching the center of the universe;it is impossible for us to rotate the cosmos as a single solid body. But thecases are on a par, introduced and appealed to in the same fashion and inthe same terms by mediæval physicists, and that strongly suggests the so-called “empirical and observational character” of some examples is simplyaccidental. Heytesbury uses a nice phrase: he says that he is proceeding se-cundum imaginationem, “according to imagination” (fol. 43vb of the Venice1494 incunabulum; see also fol. 161vb). In fact, Heytesbury explicitly statesthat some of his cases are not physically possible—acceleration to infin-ity, diminution to zero quantity—but that they are imaginable, and henceshould be considered.20

20 Both examples come from the Regulae: in the first case, a body continuously increasesin acceleration such that “it has every imaginable degree of speed all the way up to

an infinite degree of speed secundum imaginationem” (fol. 43vb); in the second case,

a magnitude is diminished incrementally at a constant rate until there is no quan-tity, and “although this is not possible strictly speaking [i. e. not physically possible],

still it may well be permitted for the sake of the argument, since it does not include

a contradiction” (fol. 48va: quamvis enim hoc not sit possibile de virtute sermonistamen ex quo casus non claudit contradictionem satis poterit admitti gratia disputa-

tionis). These passages, and the one cited in the next paragraph, are discussed inEdith Sylla, “The Oxford Calculators” in The Cambridge History of Later Medieval

Philosophy 557–560; she is following work done by John Murdoch, who, more than

any other scholar, has called attention to the secundum imaginationem procedure ofmediæval physicists. See especially John Murdoch, “Philosophy and the Enterprise

of Science in the Later Middle Ages” in The Interaction Between Science and Philos-

ophy, ed. Y. Elkana, London 1974, 64–70, and John Murdoch and Edith Sylla, “TheScience of Motion” in Science in the Middle Ages, ed. David Lindberg, Wisconsin

1978, 246–247. The ‘imaginary’ character of Heytesbury’s work is also discussed by

Curtis Wilson, op. cit. 24–25. Gaetano di Thiene, whose commentary on Heytesbury’sworks is printed with the text in the 1494 Venice incunabulum, describes the method

concisely: “even though the posited cases are not naturally possible, nevertheless they

are imaginable without contradiction, whence they are to be granted by the logician”(fol. 89ra commenting on the Sophismata: licet casus nunc positi de facto non sunt

possibiles, sunt tamen imaginabiles absque contradictione, quare a logico admittendi ;see also his remark on fol. 48va). Swineshead, too, remarks of several cases that “all

of these are conceded imaginarily and denied de facto” (see Sylla, art. cit. 562; Liber


4. THE METHOD OF MEDIAEVAL SCIENCE: THOUGHT-EXPERIMENT 13

Focus on such cases postulated secundum imaginationem: they seem tocover a wide variety of examples. They include possible situations whichwe could bring about, though there is no evidence that anyone ever did:sharpening lances after Buridan’s lecture. They also include possible situa-tions which we cannot bring about, due to our lack of the relevant power,but which could be brought about through the action of some other agent(typically God): rotating the cosmos as a single solid body. Then there arepossible situations which not even God can bring about, for example thosewhich would involve changing the past. Finally, there are situations whichare not possible at all, examples constructed on the basis of per impossibilereasoning: Heytesbury’s case of a body with actually infinite acceleration,or a body containing neither hot nor cold thence able to be simultaneouslyheated and cooled.21 The only unifying mark all these cases have, I believe,is that they are thought-experiments. In order that this not be a vacuousclaim, we shall have to look a little more closely at thought-experiments.

4. The Method of Mediaeval Science: Thought-Experiment

A more adequate picture of the scientific activities of the fourteenth cen-tury can be constructed, it seems to me, by examining the Aristotelianparadigm of scientia. Very roughly, an organized body of knowledge countedas scientia if it could be organized and understood as a deductive structure:a certain class of universal propositions are self-justifying and necessary, andtheir logical closure under the four forms of entailment (the four causes) in-cludes all other propositions of the scientia. This view gives a paradigmof knowledge, representing its ideal form: it is not meant to explain theacquisition of knowledge, but to characterize its final shape—much like thecovering-law model of scientific explanation. While there were philosophicaldisagreements on particular points, e. g. how one scientia was to be set offfrom another, the central points of the paradigm were generally accepted.

How are the basic propositions—the ‘principles’—of a scientia learned?Aristotle offered only sketchy remarks in Posterior Analytics 2.19 to theeffect that the mind, after promping through âpagwg¨, rose to grasp prin-

de calculationibus, fol. 15ra of the 1520 Venice incunabulum).

21 Heytesbury mentions this case in his Sophismata at fol. 170va, and explicitly describesit as impossible: dato per impossibile quod esset aliquod corpus alterabile quod nullam

caliditatem nec frigiditatem haberet. See Sylla, art. cit. 560, who says of this case that

Heytesbury “is interested in performing thought-experiments, but he is unconcernedwith even their theoretical realisability.” Similar cases are discussed by Swineshead

(see the preceding note).



ciples through noÜj.22 Since such principles are self-justifying, carryingtheir justificatory warrant on their face, as it were, the problem was un-derstood to be how to bring about the insight which allowed one to “see”such ineluctable truths. How does the speculative intellect get put into highgear? The answer I want to propose is that one does so through consid-ering thought-experiments, that thought-experiment is the methodology ofmediæval science.

Striking as this claim may seem, it is a corollary of a more basic mediævalview, namely that any knowledge worthy of the name of scientia is academic,the fruit of the academic method of inquiry—disputation. Siger of Brabant,for example, argues that the knowledge of truth presupposes the ability toresolve any objections, doubts, or counterexamples raised against the viewheld to be true: “the knowledge of the truth in any subject is [!!] the solu-tion of doubts; just as it is said of judges that the judging is improved byhearing arguments from both sides, so too considering first the argumentsfor each side of a contradiction, leading to doubts, improves the judging oftruth.”23 Learning how to generate objections, doubts, and counterexam-

22 This view is not explicitly stated, but is inferred on the basis of two passages in Pos-terior Analytics 2.19: at 100b2–3 Aristotle writes that “we must become familiar withthe first [principles] by induction (âpagwg¨),” and at 100b8–9 he writes “Since noth-

ing can be truer than scientific knowledge (âpist mh) except noÜj, it must be noÜj

which grasps the first principles.” The standard translation of the Posterior Analyticsduring the Middle Ages, erroneously attributed to Boethius (e. g. in Patrologia latina

64, J.-P. Migne (ed.), Paris 1847, 712D–761B), was in fact made by James of Venice,

q. v. Lorenzo Minio-Paluello, “Iacobus Veneticus Grecus: Canonist and Translatorof Aristotle,” Traditio 8 (1952), 265–304; in his translation the cited passages read:

Manifestum est quoniam nobis prima inductione cognoscere necessarium est, and

Nihil verius contingit esse scientiae quam intellectus; intellectus utique erit princip-iorum. . . (from Aristoteles latinus 4.1–4, Lorenzo Minio-Paluello (ed.), Bruges-Paris,

Desclee de Brouwer 1968, 106.14–15 and 107.1–2 respectively). Neither the transla-tions nor the original Greek force the construal of noÜj or intellectus as some form

of “intuition,” that is, as an independent faculty through which knowledge is gained.

(There is indeed some doubt that this interpretation, though standard, is correctfor Aristotle: see Aristotle’s Posterior Analytics, translated with notes by Jonathan

Barnes, Clarendon Press 1975, 256–257.) However, the standard mediæval intepreta-

tion of the passages took intellectus as a source of non-discursive knowledge. RobertGrosseteste, author of the standard commentary on the Posterior Analytics during

the Middle Ages, glosses these passages by describing the “intellectual vision” of the

first principles: the mentis aspectus is directed upon them (Robertus Grosseteste:Commentarius in Posteriorum Analyticorum libros, Pietro Rossi (ed.), Florence 1981,

406.71). Any mediæval philosopher reading the text would certainly have had Gros-

seteste’s commentary present to hand. For a discussion of the mediæval understandingof inductio, see §5 below.

23 That is, Cognitio enim veritatis in aliqua rerum solutio est dubitatorum. This claim


4. THE METHOD OF MEDIAEVAL SCIENCE: THOUGHT-EXPERIMENT 15

ples was the lifeblood of the mediæval university, and thought-experimentsare an obvious source of difficult cases.

This view may gain some plausibility if we also recall what countedas a scientia: the paradigm cases were metaphysics, mathematics, andphysics. They shared a unity of formal structure as scientiae; hence itis not unreasonable to think that they shared a unity of method as well.Thought-experiments seem peculiarly appropriate to geometry (the mostsophisticated mathematics in the Middle Ages), perhaps due to the ide-ality of geometrical objects: dimensionless points, perfect circles, and thelike. Geometric proof seems no more than an extended thought-experiment.Metaphysics, too, is full of thought-experiments, perhaps due to its a prioristructure. For example, Duns Scotus, while discussing the relation betweentime and modality, puts forward a case in which which an angel with freewill exists only for an instant, and chooses one of two exclusive options; thisimagined case allows him to argue for his view of the relation between modal-ity and time. Hence physics, like mathematics and metaphysics, should becharacterized by self-justifying basic propositions, and the speculative in-tellect is goaded into action to see the truth of these principles throughconsidering appropriately constructed thought-experiments.

A few words about thought-experiments are therefore in order. First,some take the ‘mentalistic’ aspect of thought-experiments seriously, andhence investigate topics such as conceivability and its relation to possi-bility; the nature, scope, and reliability of introspection; the mental me-chanics involved in imagination; and the like. While such topics may beworthy of investigation, they seem never to have been raised in connectionwith thought-experiments during the High Middle Ages. This suggests thatmediæval philosophers took thought-experiments to be rather like consis-tent sets of sentences, enabling them to bypass ‘mentalistic’ aspects andfocus on logical aspects. Second, I take the ‘experiment’ part of “thought-experiment” seriously: the imagined situation may be distinguished fromthe claims about what takes place in the imagined situation. It is perfectlypossible to imagine, with Swineshead, that a rod of uniform density is ap-

comes from Siger of Brabant’s preface to his questions on the Liber de causis, only re-

cently recovered, edited by A. Marlasca in Les “Quaestiones super librum De causis” deSiger de Brabant, Louvain 1972, 35. Similar sentiments are expressed by (for example)

Henry of Brussells; see Martin Grabmann, “Die Aristoteleskommentare des Heinrich

von Brussel und der Einfluss Alberts des Grossen auf die mittelalterliche Aristotele-serklarung” in Sitzungsberichte der Bayerischen Akademie der Wissenschaften, Heft

10 (1943), 82. These passages are noted and the disputational character of mediævaleducation discussed in Anthony Kenny and Jan Pinborg, “Medieval Philosophical

Literature” in The Cambridge History of Later Medieval Philosophy 11–42.



proaching the center of the universe through empty space, and to inquireafter its behavior—and indeed to disagree with Swineshead’s claims aboutwhat would transpire.

Taken together, these two considerations suggest a general characteriza-tion: a thought-experiment consists of a set of sentences D which are tobe understood as a description of a situation, together with the claim thatD describes a case in which ϕ, where ϕ describes “what happens” in thesituation.24 We may treat thought-experiments solely on the logical levelby taking p to be the sentence describing the characteristic ϕ, so that (forexample) Swineshead may be understood as giving a description D, a set ofsentences describing the case in which a rod of uniform density approachesthe center of the universe through the void, and then proposing that in anon-trivial sense p, a sentence describing the behavior of the rod, “followsfrom” D. No loss of generality is involved in the linguistic move.

These points made, the obvious question remains: how does one deter-mine what happens in a thought-experiment, which sentences follow fromwhich descriptions—and indeed what sense of ‘following from’ is relevanthere? Mediæval philosophers took the question seriously, and consciouslyaddressed the philosophy of their scientific method. The mediæval writingson obligationes are the philosophical reflections on their practice of con-structing thought-experiments and arguing on their basis; obligationes arethe metamethodology of mediæval science.

5. Obligationes

A typical obligatio25 has the formal characteristics of a debate, that is,of a mediæval disputation: there are two parties, an opponent and a re-

24 This characterization is adapted from Aron Edidin, “Philosophy: Just Like Sci-

ence Only Different,” Philosophy and Phenomenological Research 45 (1985), 537–552.Edidin does not use it as a characterization of thought-experiment, but as the form of

‘philosophical intuitions’ when presented with cases for consideration, which he devel-

ops as a candidate for contemporary metaphilosophy; he does not mention mediævalphilosophy, but much of his discussion is easily adapted for our purposes.

25 The description of obligationes is based primarily on Walter Burleigh; see Romuald

Green, The Logical Treatise ‘De obligationibus’: An Introduction with Critical Textsof William of Sherwood and Walter Burley, The Franciscan Institute 1982. It seems

likely that the treatise Green here attributes to William of Sherwood is in fact by

Burleigh as well: see Paul Spade and Eleonore Stump, “Walter Burley and the Obli-gationes Attributed to William of Sherwood,” History and Philosophy of Logic 4

(1983), 9–26. The best general introduction to obligationes is given by Paul Spadeand Eleonore Stump, “Obligations,” in The Cambridge History of Later Medieval

Philosophy 315–341.


5. OBLIGATIONES 17

spondent. The opponent begins by laying down some claim, a propositionsuch as “Socrates is running.” His action is called positio, positing, andwhat he posits is the casus, the case, also known as the positum.26 Therespondent admits the case—or, if he does not, there is no dispute—andthen the opponent proceeds to put forward (to propose) other propositions.To each proposition, the respondent either concedes, denies, or ‘doubts’ itstruth; he is ‘obliged’ to give certain responses in accord with rules; hencethe name obligationes. The point of the exercise, if there is an identifiablepoint, is to trap the respondent in a contradiction; an obligational disputeexplores “what happens” given the positum.27

Obviously, much turns on the rules according to which proposed sen-tences are conceded, denied, or doubted. And this is the material whichmakes up the bulk of the literature on obligationes. Some fairly obviousrules can be stated: any proposition which is a direct consequence of thepositum must be assented to, and any proposition the negation of which is adirect consequence of the positum must be denied. Beyond that, there is nogeneral agreement about the correct rules. For example, it seems a minimalcondition that no one dispute includes a conceding by the respondent of asentence and of its negation, whether at the same stage or different stages.But it is less clear that the same response should be given to a sentence nomatter when it is proposed. For example, a proposed sentence which wasdoubted at an earlier stage might, when proposed again, be conceded, dueto additional information gained in the course of the dispute. Of the manyphilosophical debates over the proper rules of obligationes, three in partic-ular show that obligationes constitute a theory of thought-experiments.

First, some philosophers argued that the positum need not be possible.This corresponds, in thought-experiments, to reasoning per impossibile. A

26 This is a deliberate simplification. Technically, the casus was a stipulation outside

the context of the disputation about the nature of reality, while the positum was anelement in the disputation which may or may not conform to the casus as described.

In similar fashion, and for the same reasons, the description of the start of an obli-

gational disputation is simplified: no account is being taken here of depositio anddubitatio, in which the positum is held to be, respectively, false and doubtful. This

last simplification dominates my discussion of the rules of obligationes, since it was

an open philosophical question whether the rules for depositio ought to mirror thosefor positio, to say nothing of what the proper rules for dubitatio should be.

27 That obligationes explore what “happens” given the positum is explicitly stated inthe early literature: the Tractatus Emmeranus de falsi positione, L. M. De Rijk (ed.),

“Some Thirteenth-Century Tracts on the Game of Obligation,” Vivarium 12 (1974),103.22–23, says that when falsehoods are posited they should be conceded to see what

happens—indigent positione ut concedantur et videatur quid inde accidat.



lively debate ensued in the obligationes-literature whether anything whatso-ever followed from a contradiction; the general consensus seemed to be thatthe spread of a contradiction is limited, although no precise criteria couldbe stated; many authors put severe restrictions on their use.28 Regardlessof how this rule was applied, though, it seems that ordinary propositionallogic was a component of obligational disputes: disjunction introduction,conjunction simplification, detachment—in short, all the standard rules ofpropositional logic, enhanced by mediæval logical and semantic theory—governed the relation between sentences at different stages of an obliga-tional dispute. Hence the standard practice of mediæval physicists, andindeed the standards of intellectual rigor in general, could be subsumedunder obligationes. For example, the standard practice of deducing conse-quences of a theory as a check on the acceptability of the theory falls underpropositional logic; thus reductio ad absurdum, appealing to logical incom-patibility, as well as Buridan’s three examples, appealing to commonly-heldbeliefs, can be understood in obligational terms.

Second, there were philosophical debates over how the respondent shouldtreat proposed sentences which are not logical consequences of the set of sen-tences composed of the positum and the sentences such that each was eitherconceded or its negation denied. Such sentences were called impertinens, in-dependent, and a debate among Walter Burleigh, Richard Kilvington, andRichard Swineshead arose as to their proper treatment.29 The standard

28 William of Ockham, in his Summa logicae (eds. Philotheus Boehner, Gedeon Gal,

Steven Brown, Guillielmi de Ockham opera philosophica I, Franciscan Institute Press,1974), 739 offers two restrictions: first, not all the rules for positio hold when the

positum is impossible; second, only a sentence whose impossibility is not obvious can

be laid down as a positum. The second restriction rationalizes the first, since it is clearthat the obligational dispute beginning with an obviously impossible sentence would

be short and uninteresting. Nevertheless, such cases have to be taken into account,

because otherwise the legitimacy of proof by reductio ad absurdumwould be open toquestion.

29 Walter Burleigh, in Green op. cit. §§3.14–3.16; Richard Kilvington, in his Sophismata(forthcoming, eds. N. Kretzmann and B. Kretzmann); Richard Swineshead—who is

not to be confused with Roger Swineshead, mentioned above—in his Obligationes,

edited by Paul Spade, “Roger Swyneshed’s Obligationes: Edition and Comments,”Archives d’histoire doctrinale et litteraire du moyen age 44 (1977), 243–285. During

the middle of the fourteenth century, Robert Fland described two currents of thoughtin the literature on obligationes, the ‘old view’ associated with Burleigh, and the ‘new

response’ associated with Richard Swineshead (ed. Paul Spade, “Robert Fland’s Obli-

gationes: an Edition,” Mediæval Studies 42 (1980), 41–60); Kilvington’s position isbest seen as intermediate between Burleigh and Swineshead. For simplicity I con-

centrate on Burleigh’s Rule. Kilvington’s position is not clear, but he argued that


5. OBLIGATIONES 19

view, which we may call “Burleigh’s Rule,” held that such sentences shouldbe treated as they are in the actual world: if known to be true, conceded; ifknown to be false, denied; otherwise doubted. To see why Burleigh’s Rulewas adopted, note that an obligational dispute may be viewed as askingwhat would happen were a given condition, specified in the positum, to ob-tain. What sort of reasoning takes contingent facts in the actual world asrelevant to determining what would happen under a given condition?

The answer: counterfactual reasoning, taking ‘counterfactual’ in its broadmodern sense as any subjunctive conditional. When the positum is not im-possible, each stage in an obligational dispute can be seen as constructingand narrowing down a class of possible worlds as similar to the actual worldas is compatible with what has been conceded and denied at earlier stages.30

This interpretation may be stated more exactly: let D stand for the setcomposed of the positum and sentences either conceded or sentences whosenegation was denied at any earlier stage of an obligational dispute, and let pbe a newly-proposed sentence; then to this stage of the obligational disputethere corresponds a counterfactual of the form “If D were the case then pwould be the case.” This view is confirmed by noting that, so construed,obligationes share most of the properties commonly associated with coun-terfactuals.31 Hence obligationes include the logic of counterfactuals, just

the order of the proposed sentences should make no difference to the response given

(Sophismata S47). Swineshead went farther and argued that any proposed sentenceshould be evaluated only in light of the positum, disregarding all other sentences

(ed. cit. §2). Both Kilvington’s and Swineshead’s positions allow for inconsistency

of a fairly strong sort: in the course of an obligational dispute, the same proposedsentence may be conceded at one stage and denied at another stage. Burleigh’s Rule

does not permit this, and seems to have been a far more popular view. Burleigh’s

Rule is discussed in detail by Eleonore Stump, “The Logic of Disputation in WalterBurley’s Theory of Obligations,” Synthese 63 (1985), 355–374.

30 The “counterfactual interpretation” of —it obligationes has its origins in a lecture de-

livered by Norman Kretzmann at the NEH Summer Institute on Mediæval Logic, held

at Cornell University in 1980, and has received definitive formulation in Paul Spade,“Three Theories of Obligationes: Burley, Kilvington, and Swyneshed on Counterfac-

tual Reasoning,” History and Philosophy of Logic 4 (1981), 89–120. I am indebted to

both for my understanding of the counterfactual interpretation, and rely on Spade’sexposition for my discussion.

31 For example, (i) strengthening the antecedent fails; (ii) transitivity fails; (iii) contra-

position fails—each of which is characteristic of the logic of counterfactuals. Further-more, assuming omniscience on the part of the respondent (so that dubitatio drops

out), (iv) conditional excluded middle holds, which is a part of some modern sys-

tems of counterfactuals. However, on the most obvious reading of Burleigh’s Rule,strengthening the consequent also fails, which is not a desireable feature. See Spade,

art. cit. 100–103.



as they include propositional logic. Yet the logic of counterfactuals is notthe whole of obligationes, as has been claimed, for, aside from the difficultieswith impossible positio (corresponding to per impossibile forms of reason-ing) noted above, it is a substantive claim that the distinction between theabstract and the concrete matches the distinction between the possible andthe actual. While it might be true that all possibilia are abstracta, theconverse surely does not hold; abstract entities need not be possible ‘be-ings’ in any ordinary sense. The ideal objects of geometry, for example,need not exist in any possible world. This is perhaps most clear in thecase of metaphysical speculation. To take a standard mediæval example,based on Aristotle, Metaphysics 7.3 (1029a10–20): imagine two individualsof the same species, such as Socrates and Plato, and strip away from eachall non-essential proprties; once the level of the essence has been reached,how many essences are there—one or two? This is a perfectly valid thought-experiment and was often stated in the language of obligationes, but thereis no possible world in which the mere essences of things, unencumbered bytheir non-essential properties, exist.32 Therefore obligationes include thelogic of counterfactuals, but only as a proper part.

Third, several authors allow for distinct species of positio, in cases inwhich the positum includes explicit semantic content. These varieties wereuseful in investigating logical and linguistic issues. One example of this is thekind known as institutio or impositio, in which a new meaning is stipulatedfor an expression; the question is what logical relations are preserved, e. g.“ ‘God’ stands for Brunellus the Ass” or “The sentence uttered by Socratesis false.” Other forms include explicit reference to the respondent: for thekind called petitio, the respondent is said to concede, deny, or doubt someclaim, e. g. “You concede that the King is in Paris”; for the kind called sitverum, propositional attitudes, typically stipulating the state of knowledgeof the respondent, are included, e. g. “You know that the King is seatedor not seated.” Each of these three kinds of positio explicitly involve theexpressive form of the sentences initially used to describe the situation inquestion, allowing for an investigation of the paradoxes of self-reference,

32 The latter claim is made explicitly by several authors, such as Walter of Mortagne

and Peter Abailard, who argue that such unencumbered essences could not exist; theyare careful to use the subjunctive formulation. There are further difficulties if one is

wedded to the modern understanding of possible worlds. The general difficulty is that

modern modal systems are not strong enough to distinguish between the possibilityof something and its possible being, since the possibility of something just is its being

in some possible world. This makes nonsense of Anselm, for instance; see Peter King,“Anselm’s Intentional Argument,” History of Philosophy Quarterly 1 (1984) 147–166,

in which this issue is discussed at greater length.


6. OBLIGATIONES AND METAMETHODOLOGY 21

substitutivity in semantically opaque contexts, the function of indexicalsand terms sensitive to semantic context, epistemic and doxastic logic, andthe like.

The general conclusion I want to draw should be apparent: obligationesare the mediæval theory of thought-experiments, including various forms ofreasoning which at first glance appear to have little to do with one another.Construing a thought-experiment as a set of sentences D that describe asituation, and a proposed sentence p said to follow from D, then the concernwith other forms of positio is readily intelligible. While it is usually conve-nient to talk directly of the imagined situation, using the material mode,no harm is done by retreating to the formal mode to talk about the sen-tences describing the situation, and this allows philosophical questions tobe raised about the description itself: questions involved in the philosophyof logic and language. Furthermore, when the positum is possible, the wayin which the proposed sentence is said to “follow” from the earlier stages ofan obligational dispute is clear; it or its negation either follows deductivelyor counterfactually. Finally, impossible positio takes account of per impos-sibile reasoning, and justifies methods of indirect proof such as reductio adabsurdum.

6. Obligationes and Metamethodology

To return to the question raised in §2: under what conception of ‘physics’is the Mean-Speed Theorem naturally grouped with logical and semanticpuzzles such as the Liar, problems about knowledge and belief, anaphoricreference, propositions evaluated with repect to instants and limits, func-tional extremes, and the motion of bodies in space? The answer: a concep-tion of physics as an aristotelian scientia which proceeds by way of thought-experiment. The whole battery of concepts traditionally associated withmodern science is simply absent: testing, experimentation, confirmation,induction, statistical projection, repeatability, and the like. If obligationesare part of the metatheory of philosophical method, then we should expectto see in its armamentarium features relevant to various kinds of philosoph-ical reasoning: and that is just what it provides.

We can distinguish two versions of this conclusion, however. The weakversion is that the theory of obligationes provides the logical basis for ametamethodology; the strong version is that it was consciously understoodas providing the logical basis for a metamethodology. The weak versionis undoubtedly true: the uniformity of the terminology of obligationes in



mediæval philosophical literature, and especially scientific literature;33 thecharacter of the philosophical debates over the proper rules of obligationes;the indifference to testing and experiment combined with an almost obses-sive use of counterexample, distinction, and argument—all these testify tothe truth of the weak version. And, in a sense, this is sufficient for my the-sis; characterizing a paradigm in which intellectual activity took place neednot depend on the contemporary understanding of the paradigm; there isindeed some reason to think that we, at a historical remove and free fromits influence, can more accurately describe the paradigm than a mediævalphysicist could.

The mediæval understanding of induction (âpagwg¨ or inductio), whichprompts the intellect to grasp first principles, offers further support. Aristo-tle describes induction at the start of the Topics as beginning from ândoc�,that is, beliefs that are “plausible” in the sense that they are held by every-one, or by the majority, or by the wise (either all of the wise, the majority,or the most famous).34 What is more, Aristotle explicitly says that dialecti-cal reasoning is the way in which the first principles are learned (Topics 1.2

33 Mediæval philosophical literature of this period is of two kinds: text-directed, such as acommentary, gloss, or paraphrase/abbreviation, and independent treatise. There seemto be four major styles of philosophical writing: exegesis, questions, disputations, andsophismata. (There are exceptions: opuscula such as Aquinas’s De ente et essentia;

letters and sermons; dialogues; works with literary formats. But most philosophical

literature of the fourteenth century is in one of the four major styles.) Commentaries,even those highly bound to the text, include both exegesis and questions, and it is

common to find a commentary consisting only of questions, such as commentaries onPeter Lombard’s Sententiae. Independent treatises tend to be handbooks or summae,

which often consist in quaestiones (e. g. Aquinas’s Summa theologiae); recorded dispu-

tations, either as quaestiones disputatae or quaestiones quodlibetales; or sophismata.The simple structure of each is as follows: a quaestio begins with a question, then

cites arguments, quotations, or examples supporting positive and negative answers to

the question, after which the issues underlying the question are discussed, distinctionsare drawn, and an answer endorsed, closing with the resolution of the earlier mate-

rial supporting the other answer to the question; a disputation has much the same

form, except that it begins with a thesis and the issues are treated at greater lengthand evolved in a systematic way; sophismata begin with a sentence (the “sophisma-

sentence” in Kretzmann’s terminology) and a stipulation about the case, followed by

plausible arguments both for the truth and the falsity of the sentence, after which theissues are discussed, distinctions drawn, a resolution proposed, and one set of argu-

ments replied to. The internal differences in structure, while real, nevertheless do notaffect the claim that each of these is rationalized by the theory of obligationes, since

they share the procedural feature of argument, counterexample, refutation, proof, and

the like.34 The passages in question are Topics 1.8 103b3–7 and 1.13 105a35–b3, the latter read-

ing in Boethius’s translation (standard in the Middle Ages): aut omnium opiniones


6. OBLIGATIONES AND METAMETHODOLOGY 23

101a37–b3): “since the principles [of a science] are primary to all else, theymust be arrived at in each instance with regard to the commonly-held beliefs(ândoc�); this process belongs especially or most fittingly to dialectic.” Themediæval theory of dialectic is given in treatises de obligationibus. Hencethe method involved in the intellectual grasp of first principles is systemat-ically explored in obligationes, which therefore are the metamethodology ofmediæval science.

While the weak version is not in doubt, direct evidence for the strongversion is harder to find; treatises de obligationibus rarely draw connectionsto philosophical practice; practicing physicists rarely draw attention to theirmethod. Until further research is accomplished, I can only offer some gen-eral considerations in favor of the strong version. When discussing examplesused by other philosophers or their treatment of counterexamples, mediævalphilosophers often have explicit recourse to the rules of obligationes, thoughnot at the methodological level. The many members of Merton College whowrote on obligationes never saw fit to draw the connection between thesetreatises and their research in physics.35 Furthermore, throughout the his-tory of treatises de obligationibus no explicit statement of their philosophicalrole ever appears, a fact which has lead to scholarly confusion and doubtsabout their purpose and function; some scholars have tried to infer the pur-pose of obligationes from examining their historical origins.36 Yet it would

proponenti aut plurium aut sapientum et horum vel omnium vel plurimorum vel notis-simorum (ed. Lorenzo Minio-Paluello, Aristoteles latinus 5.1–3, Bruges-Paris, Desclee

de Brouwer 1969, 19.19–21). This is precisely the sense in which Buridan appeals to

experience: he relies on beliefs which are held by everyone or by the majority (“whateverybody knows”), with recourse to the opinions of philosophical authorities (the

“wise”).

35 See the introduction to Green op. cit. for a list and discussion of the Mertonians who

wrote obligationes.

36 Aside from a logic of counterfactuals, discussed above, obligationes have been called

an all-purpose philosophical tool by Eleonore Stump in “Roger Swyneshed’s Theory

of Obligations,” Medioevo 7 (1981), 169–174; as containing “a nucleus of rules for anaxiomatic method,” according to Philotheus Boehner, Medieval Logic: An Outline of

its Development from 1250 to c. 1400, Manchester 1952, 14; as “ingenious exercises

among schoolboys [having] little objective value,” in the words of James A. Weisheipl,“Developments in the Arts Curriculum at Oxford in the Early Fourteenth Century,”

Mediæval Studies 28 (1966), 163–164, who is followed in this view by Green and

De Rijk; as a method of testing logical skills, according to Alan Perreiah, “WhatObligations Really Are,” Medioevo 5 (1979), 123; as a rudimentary theory of dialectic,

or perhaps of dialogue-logic, by Eleonore Stump, “The Logic of Disputation in WalterBurley’s Treatise on Obligations” op. cit., and by C. L. Hamblin, Fallacies, London

1970; and so on.



be a mistake to infer from their historical origins that in the fourteenthcentury they were not consciously understood as the metamethodology ofscientia. Even at the very beginning of their history treatises on obliga-tiones were closely connected with the study of fallacies, insolubles, andsophisms; the earliest treatises date from the first decades of the thirteenthcentury, that is, when translations of the Physics began to be circulated andstudied in earnest.37 It is not until the first half of the fourteenth centurythat the number of such writings begins to increase, at which time they in-crease dramatically. I think that the proper explanation of this fact is thatobligationes, as a metamethodology, were somewhat of a ‘found tool.’ Theydeveloped out of the logical milieu of the late twelfth century, when philoso-phers were trying to assimilate the Topics and the De sophisticis elenchi andcombine their understanding of these texts with the native tradition in logicwhich had evolved in the preceding century. Historically this took place asthe medieval universities were being organized, and it readily became ap-parent, with the academic monopolization of knowledge, that obligationeswould rationalize this development. The treatises on obligationes bear thestamp of their origin, but were used in the fourteenth century as the generalsupport for mediæval philosophical and scientific practice.

7. Conclusion

The Duhemian tradition claims that the components of modern scienceare found or prefigured in earlier mediæval scientific writings: a claim which,at the very least, grossly distorts the facts. There are features of this claimI would not dispute; there is a genuine similarity between mediæval andmodern authors in the vocabulary used and concepts at issue; if problemscan survive radical changes in scientific paradigm, I am willing to concedethat there is a continuity of problems as well—free fall, the nature of motionand speed, the analysis of force and resistance. But these similarities shouldnot conceal the deep divergence between mediæval scientia and modern sci-entific method. I hope to have indicated some of the complexities on the

37 Translations of the Physics appeared as early as ca. 1150, but it was only in the

first quarter of the thirteenth century that serious efforts at philosophical assimilationtook place. Only four treatises on obligationes are known to date from the thirteenth

century, and more precise dating is a matter of scholarly dispute: (i) the Obligationes

Parisienses; (ii) the Tractatus Sorbonnensis de petitionibus contrariorum; (iii) theTractatus Emmeranus de falsi positione; (iv) the Tractatus Emmeranus de impossibili

positione. Each has been edited by L. M. De Rijk in a series of articles entitled “SomeThirteenth-Century Tracts on the Game of Obligations,” Vivarium 12 (1974) 94–123,

13 (1975) 22-54, 14 (1976) 26–49.


7. CONCLUSION 25

mediæval side; I would expect a closer look at Galileo to show that hisprocedure was more ‘mediæval’ in this respect than has generally been ac-knowledged. To some extent these are non-issues: who anticipated whomis neither historically pressing nor philosophically interesting. But this dis-missal conceals the deeper point of the nature of the mediæval scientificparadigm and its historical relation to the beginnings of modern science,which is, I take it, both historically pressing and philosophically interest-ing.

The method of thought-experiment has virtues as well as vices. It is pecu-liarly well-suited for uncovering conceptual incoherencies and inadequacies;it demands a high degree of rigor as well as logical sophistication; it is asprecise an analytic tool as can be found. More generally, it seems appro-priate for investigating a priori truths, such as those found in mathematicsand geometry, and perhaps physics. Its vices, especially as a method forscientific knowledge, are only too apparent. While it is a method whichsupports armchair speculation, there is only so much that can be learned inan armchair, without engaging in concrete practice and activity. Thought-experiments, in their mediæval use, support theories which have no checkor control, no way to test their correctness or incorrectness, as opposed tothe modern experimental method. In the pejorative sense associated withmediæval philosophy ever since the Renaissance, any dispute over the cor-rectness or incorrectness of a theory which is isolated from practice is apurely scholastic question.


mediaeval thought-experiments: the metamethodology...

Documents